Hossein-Bor, Amineh (2023) Comparison of some digital soil mapping methods in spatial modeling of soil erodibility factor in Sistan plain. Masters thesis, university of zabol.
|
Text
Comparison of some digital soil mapping methods in spatial modeling of soil erodibility factor in Sistan plain.pdf Download (54kB) | Preview |
Abstract
Soil erosion is one of the most important aspects of natural resource degradation. In the revised universal soil loss equation (RUSLE), the soil erodibility factor (K) is one of the main factors affecting the amount of soil loss, and its determination is essential in estimating soil erosion. Digital mapping is also an effective technique for better expression and interpretation of soil information, and it can be used to quickly and accurately prepare and update soil maps. This research was designed and implemented with the objectives of investigating the variations and spatial distribution of soil erodibility factor (K) and determining the best method of digital soil mapping in modeling the K-factor in the Sistan plain. For the spatial modeling of factor K, the variables of geomorphometry and remote sensing were used. Machine learning models, including random forest (RF) and K-nearest neighbor (k-NN), as well as geostatistical models of ordinary kriging (OK), Co-kriging (COK) and inverse distance weighted (IDW), were utilized for modeling K-factor. The results showed that the RF model predicted K-factor with an overall accuracy of 81% and a kappa index of 0.60, which was better than the k-NN method and the studied geostatistical methods. The largest predicted area belonged to the very high erosion class with 53%. Additionally, the RF model using scenario 1 had the highest accuracy for predicting K-factor with an R2 of 0.06 and an RMSE of 0.007. The auxiliary variable of valley depth (VD) had the highest importance in predicting K-factor, followed by relative slope position (RSP), which reflects the importance of water flows in the region. Overall, it appears that using the RF method along with auxiliary variables reflecting the conditions of the area (river deltas) can be highly beneficial in creating a digital map of K-factor.
| Item Type: | Thesis (Masters) |
|---|---|
| Uncontrolled Keywords: | remote sensing, geomorphometry, water erosion, Revised universal soil loss equation (RUSLE), random forest. |
| Subjects: | S Agriculture > S Agriculture (General) |
| Depositing User: | Mrs najmeh khajeh |
| Date Deposited: | 03 Dec 2024 08:16 |
| Last Modified: | 03 Dec 2024 08:16 |
| URI: | http://eprints.uoz.ac.ir/id/eprint/3719 |
Actions (login required)
 |
View Item |
